The Study of Typhoon Rainfall Forecasting in Hualien Area

• Main Authors: Hung-Yu Lai, 賴泓宇
• Other Authors: Horng-Yu Yang
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/46240777077637906140

碩士 === 中華科技大學 === 土木防災工程研究所在職專班 === 101 === Majority of Typhoons mostly land from the east side of Taiwan, and were higher rainfall in windward side. The research focuses on the Hualien area located on the east of Taiwan, and is objected to set up a rainfall prediction modeling by looking at different typhoons’ routes, then effectively predict rainfall when different typhoons occur, to reduce the loss caused by the typhoon when it occurs. This paper is to discuss how to reduce the loss caused by rainfall when typhoon occurs by enhancing the ability of rainfall prediction. This study adopts the rainfall data provided by corresponding authorities, to achieve the aims described as follows: (1) Using linear regression models to create different typhoon rainfall forecasting model. (2) The precision of the outcome that is analyzed and compared with each typhoon with similar route in the past accordingly, and (3) Comparing the prediction effects as to the different typhoon rainfall routes for future reference. The results show:1. Typhoon path 2 In terms of linear regression prediction model, R2 was 0.175 and value F was 8.934, significant functionality was 0.000 (<0.001), which shows it reaches the level of significance. 2. In Typhoon Rainfall path 3, its linear regression prediction model R2 was 0.207 and value F was 14.105, significant functionality was 0.000(<0.001), which shows it reaches the level of significance. 3. In Typhoon Rainfall path 4, its linear regression prediction model R2 was 0.294 and value F was 14.105, significant 0.000(<0.001), which shows it reaches the level of significance. 4. In the combination of Typhoon Rainfall path 2、3、4, its linear regression prediction model R2 was 0.176 and value F was 31.143, significant functionality was 0.000(<0.001), which shows it reaches the level of significance. Based on the simulation outcomes, it shows that significant functionality existing in each model. In addition, all models are based on MAPE assessment , MAPE values of path 2 and the combination of path 2、3、4 were 1.64% and 1.71%, respectively, results seem to be good. This study shows typhoon rainfall linear regression prediction model were easy to operate and the overall effects were great in general. Thus it can be further referred by the corresponding authorities in the area of typhoon rainfall forecasting.